1. Field of the Invention
The present invention relates to native metal nitride substrates for nitride deposition. The invention also relates to the growth of a nitride epitaxial layer on a non-native substrate. The invention further relates to GaN substrates having heavy n-type doping. The invention still further relates to a method of detaching a high quality nitride epitaxial layer from a silicon substrate.
2. Background of the Related Art
Gallium nitride (GaN) is emerging as an important material for high technology applications. For example, GaN is currently used in the manufacture of blue light emitting diodes, semiconductor lasers, and other opto-electronic devices. However, it is currently impossible to fabricate bulk GaN crystals of usable size as substrates in semiconductor manufacturing using prior art methods and materials. Therefore, GaN films, heretofore, have been made by deposition on a non-native substrate material, typically sapphire (Al.sub.2 0.sub.3). Due to the large lattice mismatch and thermal mismatch which exists between the A1.sub.2 0.sub.3 substrate and the GaN layer, the resulting structure has large dislocation densities, thereby limiting the performance of devices fabricated from these films.
Furthermore, sapphire substrates have many disadvantages from the point of view of electronic device manufacturing. For example, sapphire substrates are electrically insulating, so that all electrical contacts must be made from the epitaxial side, thereby increasing the difficulty and cost of device fabrication. In addition, sapphire is a poor conductor of heat, which can limit the power of devices which incorporate this material. Also, sapphire is mechanically very hard, and is thus difficult to remove from the epitaxial film. Finally, sapphire is an expensive material, which leads to increased cost of the device. For all these reasons, there is a need in the semiconductor industry for a GaN native substrate.
Substrates other than sapphire have been investigated in the prior art for GaN epitaxial deposition. However, each of these substrates suffer from their own drawbacks. For example, silicon (Si), having low cost and being widely used in the semiconductor industry, would have appeared to be an attractive material for use as a non-native GaN substrate. However, it is impossible to grow an epitaxial film of GaN directly on a silicon substrate using growth techniques of the prior art. Thus, a thin but resistant layer of amorphous silicon nitride (Si.sub.3 N.sub.4) forms preferentially on the silicon surface, during the growth process. This silicon nitride layer interferes with the ability of the underlying Si crystal to induce order in the epitaxial layer. The result is that the GaN nucleating on the nitridized surface is polycrystalline or amorphous, unlike the single-crystal layers grown on sapphire. Also, there is a thermal mismatch between the Si and the GaN layers; i.e., they have different coefficients of thermal expansion (CTE), which induces strain and cracking in the epitaxial layer upon cooling. In order for GaN layers grown on Si to be technologically useful, these problems must be overcome.
Other prior art methods of overcoming the problems associated with the use of sapphire substrates for GaN growth have involved removal of the substrate from a thick epitaxial GaN layer. For example, according to one prior art technique, the growth substrate can be completely etched away, or "sacrificed," at the growth temperature, either before or after the GaN deposition is complete. This technique, however, has several disadvantages. For example, the additional etching step adds cost and complexity to the process: controlling the two processes and ensuring that they remain properly isolated from one another is difficult. In addition, it can be difficult to retain the GaN layer in place once the sacrificial substrate is etched away. Alternative techniques have involved removing the unwanted substrate in a separate process after growth of the epitaxial layer. An example is the wafer bonding technique, which allows the substrate to be removed after the sample has been taken out of the growth system by mounting the substrate-epitaxy structure upside down and subsequently removing the substrate. This method does not resolve any thermal mismatch problem in the original substrate-epitaxial system, and again requires a separate substrate removal step.
The present invention resolves problems associated with producing high-quality substrates of a material (e.g., GaN) which cannot be fabricated by traditional bulk-crystal techniques, and which suffers from degradation of quality when deposited on an unsuitable substrate. Prior art techniques generally produce crystals of low quality on technologically inconvenient substrates, or involve separate processes to grow the epitaxial layer and remove the substrate. The present invention provides an integrated technique where growth and detachment of a GaN substrate may be achieved in the same reactor system. The result is a high-quality, single-crystal GaN substrate, heavily n-doped with silicon. This substrate is more useful for subsequent deposition of higher-quality epitaxial layers than are achievable using a non-native substrate. The material formed according to the invention is useful for applications in a wide array of electronic and opto-electronic devices, such as light-emitting diodes and semiconductor lasers.